The present invention discloses a method for performing a stochastic analysis of one or more hydrocarbon reservoir exploitation strategies taking into consideration one or more uncertain parameters. The method optimizes an objective function that considers the gain in value of a reservoir management goal attributable to these exploitation strategies. The methodology may be used to decide whether or not to implement a strategy. Alternatively, it may be used to decide which competing strategy will yield the maximum benefit. In another embodiment of the present invention, the value of information obtained from the installation of additional sensors or new measurements is also considered.
|
1. A method for managing risk associated with one or more hydrocarbon reservoir exploitation strategies used for determining one or more risk management decisions, the method comprising:
a. developing a reservoir model that incorporates a stochastic model of one or more uncertain model parameters;
b. identifying one or more goals for optimization;
c. identifying said one or more hydrocarbon reservoir exploitation strategies for achieving said one or more goals based on said reservoir model and associated uncertain parameters;
d. identifying control variables and constraints associated with each of said one or more hydrocarbon reservoir exploitation strategies;
e. identifying a set of equally probable reservoir model scenarios;
f. formulating an objective function based on said one or more goals to determine the gain in value attributable to said one or more hydrocarbon reservoir exploitation strategies; and
g. optimizing said objective function via at least one processor by using said set of scenarios and said control variables and constraints for each of said one or more hydrocarbon reservoir exploitation strategies to determine the optimum gain in value at a plurality of risk levels;
h. creating a risk reward summary to analyze the determined optimum gain in value at said plurality of risk levels corresponding to said one or more hydrocarbon reservoir exploitation strategies that identifies said one or more risk management decision to optimize reservoir production value; and
wherein said objective function is defined as
Fλ=μ−λσ wherein
such that, F is said objective function, λ is a risk aversion level, μ is a mean of the gains in value of the goal and the standard deviation, σ is the standard deviation of these gains, N is the set of equally probable scenarios, gi, is the gain for ith scenario; n is the number of discrete points in each of said one or more uncertain parameters; and m is the number of uncertain parameters.
2. The method of
3. The method of
4. The method of
i. collecting additional information;
j. determining a second set of scenarios based on said additional information; and
k. optimizing said objective function based on said second set of scenarios to determine the optimum gain at two or more risk aversion levels.
5. The method of
l. comparing the gain from (g) to the gain from (k) at two or more common preferred risk aversion levels to determine the value of said additional information.
6. The method of
i. formulating a second objective function wherein one of said set of scenarios is assumed to be deterministic;
j. deterministically optimizing said second objective function to determine the gain attributable to the deterministic treatment of said scenario;
k. determining the difference between the gain of (i) and the gain of (g) at a preferred risk level; and
l. repeating (i), (j), and (k) for each scenario.
7. The method of
m. determining the average of the differences in gains obtained in (l) to determine the value of perfect information.
8. The method of
9. The method of
|
The present invention claims priority from U.S. Provisional Application Serial No. 60/384,065 filed May 29, 2002, incorporated herein by reference in its entirety.
The present invention relates to reservoir management. More particularly, the present invention relates to a method for decision-making in reservoir risk management that includes a stochastic reservoir model and optionally accounts for the value of information.
Due to the inaccuracy of logging tools and unpredictable natural variations in geometry and geological parameters, there is considerable uncertainty as to the detailed characterization of a hydrocarbon reservoir. This uncertainty, coupled with the dramatic variations in the market value of hydrocarbon production, has heightened the importance of financial factors and risk management in reservoir strategies so as to maximize reservoir value.
Reservoir risk management involves optimizing assets given inherent reservoir uncertainties and minimizing risk by reducing these uncertainties. Optimizing assets usually involves making decisions about technologies and strategies (such as advanced completion systems, drilling a new well, setting injection or production target rates, etc.) and quantifying the value of implementing the proposed technology in the presence of physical and financial uncertainties. Physical uncertainty includes uncertainty in the type of reservoir model used and the properties used to populate the model. Financial uncertainties refer to uncertainty in the financial variables associated with the asset, such as the discount rate, hydrocarbon price, etc. The optimization process thus should be stochastic, with a risk level (or alternatively a confidence level) associated with the optimized solution. Based on the cost-benefit analysis (cost of implementing the technology versus the gain or value from the implementation) and the associated risk level, a decision may be made on implementing the technology. Minimizing the risk level involves gathering information about the reservoir to reduce or eliminate the inherent uncertainties. The cost of gathering this information should be balanced against the value the information brings to the stochastic optimization process. Valuation of information can then guide decisions on implementing monitoring technologies for uncertainty reduction.
Risk analysis has taken a much more prominent role since the late 1990's when hydrocarbon production costs significantly fluctuated. The oil and gas industry has taken a harder look at the technology available to perform more accurate and more efficient risk analysis. Bailey et al.'s “Taking a Calculated Risk,” Oilfield Review, Vol. 12, No. 3, pages 20-35 (2000) (incorporated by reference herein in its entirety) surveyed the major topics of interest to perform risk analysis at various degrees as well as the role and impact these techniques have on the operators in the petroleum industry. This was further exemplified using an extensive case study of a development offshore in Coopersmith et al.'s “Making Decisions in the Oil and Gas Industry,” Oilfield Review Vol. 12, No. 4, pages 2-9 (2000) (incorporated by reference herein in its entirety). As the oil and gas industry shifts its emphasis from profit by cost-cutting to a diversification of asset-management practices, global portfolio optimization akin to what is prevalent in the financial market industry could potentially be an extremely valuable exercise as shown by Adams et al. in “Portfolio Management for Strategic Growth,” Oilfield Review, Vol. 12, No. 4, pages 10-19 (2000) (incorporated by reference herein in its entirety).
Current industry practice for consideration of uncertainty in oilfield risk analysis typically involves multiple realizations of a given reservoir model (the level of sophistication ranging from simple spreadsheet tank [Material Balance] models to a geologically detailed full-scale model). This approach to reservoir risk analysis is not amenable to true stochastic optimization. For example, the optimum flow rates for wells in a faulted reservoir will depend strongly on the transmissibility of these faults that are often uncertain and can significantly affect the fluid flow in the reservoir. Hence, to obtain the optimum flow rates, one needs to integrate the fault transmissibility uncertainty and its effect on fluid flow into the stochastic optimization routine. This is not feasible with current treatments of uncertainty.
An optimization methodology for deterministic reservoirs (without any uncertainties in reservoir parameters) was applied in the 1980's to determine the optimum injection strategy for surfactants by optimizing the difference between gross revenue and the cost of injection chemicals. This methodology is described by Fathi et al. in “Use of Optimal Control Theory for Computing Optimal Injection Policies for Enhanced Oil Recovery,” Automatica, Vol. 22, pages 33-42 (1984) (incorporated by reference herein in its entirety) and by Ramirez in Applications of Optimal Control Theory to Enhance Oil Recovery (1987) (incorporated by reference herein in its entirety). More recently, commonly owned U.S. Ser. No. 09/930,935 to Couet et al. (the '935 application) and Burridge et al. “Optimal Stimulation of Oil Production,” Decision Making under Uncertainty: Energy and Power, IMA Volumes in Mathematics and Its Applications, Vol. 128, pages 17-37 (2002) (incorporated by reference herein in their entireties) address the extended issue of optimization of oil recovery subject to various physical uncertainties (stochastic optimization) and minimization of downside risk.
While these studies describe algorithms for stochastic optimization of reservoirs, they do not provide methodologies for formulating objective functions suitable for decision-making. As will be shown below, useful methodologies should allow for decisions based on a cost-benefit analysis for a technology implementation or choosing between technologies in the presence of uncertainties. This requires direct optimization of the difference between the value of the management goal (such as net present value, hydrocarbon production etc.) obtained when the technology is implemented and the value of the goal obtained for a reference case without that technology. The optimization of this difference or gain in value of a goal in accordance with the present invention is in contrast to the traditional approach that only considers the difference between the individually optimized values of the goal for the two cases. The results can be very different for the two approaches as values of the goal obtained for both the technology and reference cases are dependent on the inherent reservoir uncertainties. Accordingly, there is a need for a method that accounts for the affect of inherent reservoir uncertainties, such as in cost-benefit analysis. Furthermore, there are no existing methodologies to account for the gain in value of a goal associated with the reduction in uncertainty due to acquisition of new information, which can be defined as the value of information.
Accordingly, it is an object of the present invention to provide methods for decision-making in reservoir risk management that are stochastic and account for the gain in value attributable to one or more hydrocarbon exploitation strategies/technologies, wherein this gain in value of the goal is defined as the value of that strategy/technology.
It is yet another object of the present invention to provide decision-making methods which further account for one or multiple uncertainties in reservoir properties.
It is also an object of the present invention to provide methods that allow the quantification of the value of information in reservoir risk management.
The present invention discloses a method for performing a stochastic analysis of one or more hydrocarbon reservoir exploitation strategies taking into consideration one or more uncertain parameters. The method optimizes an objective function that considers the gain in value of a goal (such as NPV, hydrocarbon production etc.) attributable to these exploitation strategies. The methodology may be used to decide whether or not to implement a strategy. Alternatively, it may be used to decide which competing strategy will yield the maximum benefit.
In one embodiment, a method for managing risk associated with one or more hydrocarbon reservoir exploitation strategies is disclosed comprising: (a) developing a reservoir model that incorporates a stochastic model of one or more uncertain model parameters; (b) identifying one or more goals for optimization; (c) identifying one or more hydrocarbon reservoir exploitation strategies for achieving one or more goals based on the reservoir model and associated uncertain parameters; (d) identifying control variables and restraints associated with each of the one or more decision strategies; (e) determining a set of equally probable scenarios for the reservoir model (i.e., ‘N’ scenarios realizable due to the uncertainties in the reservoir parameters); (f) formulating an objective function based on the goals to determine the gain in value attributable to the hydrocarbon reservoir exploitation strategies; and (g) optimizing the objective function using the ‘N’ scenarios and the control variables and restraints for each of the one or more hydrocarbon reservoir exploitation strategies to determine the gain in value at one or more risk aversion levels. A non-limiting exemplary list of potential goals for optimization include maximizing net present value, maximizing hydrocarbon production, maximizing return on investment, minimizing cost, and maximizing hydrocarbon drainage. Likewise, a non-limiting exemplary list of hydrocarbon reservoir exploitation strategies includes installing control technology, drilling additional wells, upgrading surface facilities and performing well workovers. One skilled in the art would recognize that other goals and strategies might be appropriate in the field of hydrocarbon reservoir exploitation.
Optionally, the method of the present invention may include consideration of the value of additional information (such as that collected from additional sensors or new measurements). Accordingly, the method may further comprise determining a new set of equally probable reservoir model scenarios based on potential reduction in uncertainty attributable to the collection of additional information. The stochastic optimization is now done over this new set of ‘N’ scenarios using the previously developed objective function at one or more risk aversion levels to determine the optimum gain. In this embodiment, this optimized gain may be compared to the gain without collecting information at one or more common preferred risk levels to determine the value of the additional information. The value of the additional information may be compared to the cost of collecting the information.
Further, the method of the present invention may be adapted to determine the value of perfect information where all the uncertainty is assumed to be completely eliminated. In such a model, each of the first set of equally probable reservoir model scenarios is treated as a deterministic case and independently optimized. The difference in gains between the deterministic optimization and the stochastic optimization at a preferred risk level for each scenario would be the value of perfect information (maximum benefit of any measurement or sensor) if that scenario happened to be the true reservoir model. Because the true scenario is unknown a priori, an average of these values for all the equally probable scenarios gives an average value of perfect information for that reservoir.
For a preferred risk aversion level, λ the objective function, Fλ, may be formulated using the utility function of the '953 application, but where the mean, μ is the mean of the gains in value of the goal and the standard deviation, σ is the standard deviation of these gains as:
and where wi is 1/N; gi is the gain for scenario i. N is the number of equally probable reservoir model scenarios and is equal to nm, where n is the number of discrete points in each of the uncertain parameters and m is the number of uncertain parameters.
While this is the preferred formulation of the objective function, one skilled in the art would recognize that other stochastic objective functions (using different combinations of the mean and standard deviation) to maximize gain and minimize risk could be employed in accordance with the present invention. The objective function may further account for the cost of implementing the exploitation strategy and therefore provide a basis for a cost-benefit analysis. Further, the objective function may include a (stochastic) pricing model.
To assist in the decision-making process, the gain in values of the goal determined in the present invention may be presented in a decision table or efficient frontier for various risk aversion factors or levels.
Further details, features and embodiments of the invention will become more readily apparent from the detailed description that follows.
Reservoir risk management involves optimizing assets given inherent reservoir uncertainties and minimizing risk by reducing these uncertainties. The decision to implement any reservoir management technology is usually based on a cost-benefit analysis. Consequently, the correct valuation of value-adding technology (where the valuation could be based on net present value, hydrocarbon production, etc.) becomes a significant part of a reservoir management process. A crucial underlying factor in such an analysis is the uncertainty in the reservoir model and its properties, and in the financial variables associated with the asset. The former includes, for example, uncertainty in properties such as reservoir geometry or permeability in the reservoir, while the latter refers to uncertainty in financial parameters such as discount rate or hydrocarbon price.
The present invention discloses methods for the valuation of technology and information related to reservoir risk management. For demonstration purposes, these techniques are applied to valuation of control technologies in a reservoir with single as well as multiple uncertainties in its physical properties. Due to the presence of these uncertainties, valuation includes a stochastic optimization of one or more hydrocarbon reservoir exploitation strategies, where the objective function includes a risk aversion factor. By applying these techniques, a decision table may be generated showing the optimum gain in the value of the reservoir management goal (such as NPV, hydrocarbon production) at various confidence/risk aversion levels due to implementation of a technology. The analysis may be further extended to estimate the added value to the reservoir if the uncertainties were to be reduced (or even eliminated) by implementing appropriate monitoring technologies. This process enables quantification of the value of information (i.e., sensor value) and is useful in designing monitoring strategies to collect information.
Such techniques are applicable in cost-benefit analysis and other decision-making activities for reservoir risk management and can, therefore, be used as tools for the valuation, pricing, and marketing of technology.
The specific control technology applied in the example chosen to demonstrate the methods of the present invention is an advanced completion systems (ACS; also known as smart or intelligent completion system) as disclosed in: (1) Rundgren, et al.'s “Installation of Advanced Completions in the Oseberg 30/9-B-38B Well,” paper SPE 71677 presented at the SPE Annual Technical Conference and Exhibition, New Orleans, Sept. 30-Oct. 3, 2001; and (2) Algeroy, et al.'s, “Controlling Reservoirs from Afar,” Oilfield Review Vol. 11, No. 3, pages 18-29 (1999). These articles are incorporated by reference herein in their entireties. ACS enable control of fluid flow into selective completed zones of the well through variable choke valves that can be remotely controlled at the surface. By adjusting the valve opening, the flow or producing pressure can be set in the completion zone to any desired value. This allows the manipulation of the fluid flow in the reservoir for more efficient performance and hence an increase in the value of the goal being optimized. Note that while the “Value of Technology” examples below choose the maximization of Net Present Value (NPV) as the goal being optimized and the implementation of ACS technology as the exploitation strategy, the present invention may be equally adapted to optimizing other goals using other strategies.
Information about reservoir parameters (i.e., the uncertain parameters) may be obtained by making new measurements using surface or downhole sensors, temporarily or permanently installed, including wireline logging of wells. The cost of obtaining information from such sensors needs to be balanced against the expected benefit to the goal being optimized. Accordingly, in the example below (with respect to “Value of Information”), NPV is optimized. While the examples below choose the maximization of NPV as the goal to be optimized, the present invention may be equally adapted to optimizing other goals as well. Throughout this specification, NPV will be representative of all potential goals for optimization. The cost of collecting additional reservoir information may include not only the hardware cost but also the cost of interpreting measurements. Valuation of additional information can help determine the pricing structure for this interpretation.
The present invention uses a stochastic approach that accounts for the gain in value of a chosen goal (such as NPV in the examples below) attributable to the proposed exploitation strategy (such as ACS in the examples below). The continuous distribution of the one or more uncertain parameters may be approximated by N equally probable reservoir model scenarios, such that the mean and standard deviation of these N points is equal to the original mean and standard deviation. The gain in value of the goal chosen for optimization (such as NPV) that is due to the implementation of the exploitation strategy (such as ACS) is calculated for each of these N scenarios. The mean μ (and standard deviation σ) of the NPV gain for the stochastic case is then the weighted NPV gain (and weighted standard deviation) of the N possible scenarios:
where wi here is 1/N and gi is the gain for scenario i. N is equal to nm, where n is the number of discrete points in each uncertain parameter and m is the number of uncertain parameters. Accordingly, in situations where there is only one uncertain parameter, m=1 and N=n. Where there are two uncertain parameters, m=2 and N=n2; where there are three uncertain parameters, m=3 and N=n3, etc.
1. Valuation of Technology
As shown in
The next step is to determine the ‘N’ equally probable reservoir model scenarios 30 arising due to the ‘m’ uncertain parameters. An objective function is then formulated 35 to allow for stochastic optimization to determine the gain in the value of the goal due to an exploitation strategy at one or more risk aversion levels, preferably at one or more risk aversion levels that reflect the decision-maker's risk aversion preference (also referred to as a “preferred risk aversion level”). The function may also include a pricing model that will allow reservoir outputs to be converted to monetary units. The pricing model itself could be deterministic with fixed financial properties or stochastic with fluctuations in financial properties, such as hydrocarbon price. Optionally, a decision table or efficient frontier may be developed 40 based on the optimized objective function 35. Note that the decision table/efficient frontier may be developed for various risk aversion levels greater than or equal to zero.
The objective function should be formulated so that it provides for decision-making in the presence of uncertainties. If the exploitation strategy is installation of ACS, the benefit is the added value (or gain) to the reservoir management goal due to the implementation of that system. The objective function is hence formulated as the added value to the goal being optimized due to the implementation of this technology given the associated uncertainties in reservoir properties. Thus, for example, it could be increase in NPV from the reservoir production due to ACS implementation. For a deterministic evaluation of the benefit of the technology, the conventional practice is to maximize the NPV with ACS and compare with a reference case where there are no ACS. The difference is the added value of the technology, which can be compared to the cost of implementing the technology.
For the deterministic case, the optimization results are the same whether one optimizes for absolute value or the gain in value of the goal (such as NPV). This is not true when performing a stochastic optimization. The present invention proposes a stochastic evaluation of the technology by defining the objective function to be directly the difference between the optimum value of the goal with the exploitation strategy (for example, NPV with ACS) and the optimum value of the goal for the reference case (i.e., the NPV without ACS). The reference case is also a function of uncertainty and, hence, the gain attributable to the technology will be dependent on the uncertain variable values. By defining the objective function to be the gain in value of the goal (NPV) due to the technology implementation (ACS), the utility function proposed in the '935 application may be applied. For the example where the goal is maximizing NPV, the mean is now the mean gain in NPV and the standard deviation is the standard deviation in NPV gains. This allows the optimization algorithm to account for the fact that the NPV of both the ACS case and the reference case are dependent on uncertain parameters. The output may be a decision table (as shown below) or an efficient frontier (as described in the '935 application).
Depending on the purpose of reservoir management, the objective function could represent the added hydrocarbon production, increase in recovery, or reduction in water for the technology-implemented case as compared to a reference case without the technology. Thus, for example, hydrocarbon production maximization could be chosen as the goal to be optimized. The objective function would then be formulated based on the gain in hydrocarbon production due to ACS implementation as compared to the reference case hydrocarbon production in the absence of ACS technology. Further, one could also formulate the objective function to be the difference in value of the goal being optimized for two cases with competing strategies. For example, the objective function could be the difference in NPV when a horizontal well is being drilled as opposed to a vertical well. In this case, the reference case could be the drilling of a vertical well. Accordingly, the method of the present invention allows decision-making whereby one decision strategy serves as a reference case and the gain attributable to one or more alternative strategies are measured relative to the reference case.
In this example, the goal is to maximize this gain in NPV over a two-year period and the exploitation strategy is to install ACS. Due to the uncertainty in the physical properties, a total of two in this example, a stochastic optimization and evaluation of the reservoir is preferred to make a decision on implementing ACS technology. The reference case NPV (in the absence of ACS) depends on the width of the high permeability zone as well as on the aquifer strength. The gain in NPV also depends on these parameters. Thus, for fixed aquifer strength, the reference NPV is likely to be higher for a smaller width of the high permeability zone because of lower water cuts, but the gain from implementing the ACS technology will be lower. The reverse is true for a larger width of the high permeability zone.
For a deterministic evaluation, there is only one likely scenario and it does not matter if one optimizes for maximum NPV or gain in NPV; the result is the same. For a stochastic optimization with multiple uncertainties, there are multiple scenarios of the reservoir model with varying probabilities. Accordingly, there is a distribution of NPV's for the reference case and a different distribution of the gains in NPV due to ACS corresponding to the flow rates chosen. It is important, in doing a stochastic evaluation of implementing ACS technology, to maximize the gain in NPV, as the decision is based on the gain from implementing ACS versus the cost of the technology, rather than on the absolute value of NPV. A conventional deterministic evaluation as is practiced today as well as a more comprehensive stochastic evaluation in accordance with the present invention are provided below. Schlumberger's Eclipse™ simulator was used to model the fluid flow in the reservoir while the price of oil was taken to be constant and the interest rate was fixed in time.
Deterministic evaluation: Conventionally, a deterministic evaluation is performed in which a most likely value for the two uncertain parameters is picked and the control case is compared to a reference case as shown in Table 1 below.
TABLE 1
Cost-Benefit Analysis of ACS
Cost-Benefit Analysis
NPV without ACS
$79.2 MM
NPV with ACS
$90.6 MM
Gain due to ACS
$11.4 MM
Cost of 3 ACS
$2.25 MM
Benefit of ACS > Cost
The reference case is the case in which the technology is not implemented, i.e., its corresponding NPV is obtained by operating the well without ACS. Constraints include staying above bubble pressure and keeping the total flow rate fixed at 10,000 barrels per day (bpd). Table 1 summarizes the benefit of using ACS with optimum flow rates for a 700-ft wide high permeability zone and an aquifer PI of 4,500 bpd/psi. The gain from using this technology is the difference in NPV between the ACS case (using three advanced completion zones with optimum flow rates) and the reference case. The average cost of a single system is about $0.75 MM. Thus the gain in NPV over a 2-year period ($11.4 MM) pays back the cost of investing for the three completions ($2.25 MM) today.
The conventional way to consider uncertainty is to evaluate the scenarios at upper and lower bounds of the uncertain parameters. Thus, a ‘worst case scenario’ can be considered wherein the least gain is expected. However, this number does not indicate the probability of getting this scenario or how this gain should be incorporated with the most likely scenario to make a decision on implementing this technology. There is no information about the risk level and the gain expected or what the optimum strategy should be. If one chooses to use the optimum flow rates corresponding to the most likely case, then there is no knowledge of the downside risk. There is no knowledge of the risk involved if the uncertain parameters have values different from their medians. Furthermore, there is no way to evaluate the value of information.
Stochastic reservoir scenario—one uncertain parameter: In the preceding example, two uncertain parameters were identified. For simplicity, only one of the uncertain parameters, the width of the reservoir, will be considered here; the example below will account for concurrent consideration of both uncertain parameters. In reality, it is likely that a range of width values around a most likely value would be known rather than the exact width. This raises the question as to how to value the technology and determine if an ACS will pay back its cost, given this uncertainty in width. Based on prior knowledge from log interpretations, the variation in the width of this zone is assumed to be a normal distribution with a mean of 700 ft and a standard deviation of 327 ft. The continuous distribution may be approximated by N discrete equally probable points, such that the mean and standard deviation of these N points is equal to the original mean and standard deviation. The gain in NPV that is due to ACS implementation is calculated for each of these N scenarios. The mean μ (and standard deviation σ) of the NPV gain for the stochastic case is then the weighted NPV gain (and weighted standard deviation) of the N possible scenarios as shown in Equations (1) and (2), where there is only one uncertain parameter, m=1, so that N=n.
Once the objective function is formulated, the utility function presented in the '935 application may be used and a stochastic optimization may be performed using the weighted mean, μ, and the weighted standard deviation, σ, of the NPV gains calculated for these N scenarios,
Fλ=μ−λσ (3)
where λ is a risk aversion factor chosen from a set of preferred risk aversion levels. The objective function maximizes μ for a given σ or minimizes σ0 for a given μ. A value of λ equal to zero causes the optimization algorithm to maximize the mean without concern for the standard deviation. As the value of λ increases, the optimization algorithm puts more weight into minimizing the standard deviation; i.e., it reduces risk. If the gains in NPV of the various individual scenarios are assumed to be normally distributed, then the weighted mean corresponds to a 50% confidence level for λ equal to zero (Fλ=μ). When λ is equal to 1, the function being optimized is (Fλ=μ−1σ), and the gain is maximized at an 84% confidence level.
For the analysis here, three equally probable widths of the high permeability zone (N=3) are used. Table 2 below is the decision table that summarizes the gains achieved at various confidence levels.
TABLE 2
Decision table for cost-benefit analysis with uncertainty in the width of the high-
permeability zone.
Confidence
Flow Rates (bpd)
λ
Levela
Function Optimizedb
(f1/f2/f3)
Gain
Benefit > Costc
0
50%
F = μ − 0σ
947/743/8310
$10.37 million
Yes
1
84%
F = μ − 1σ
10/2986/7004
$7.35 million
Yes
2
97.7%
F = μ − 2σ
189/3412/6399
$5.85 million
Yes
aassuming normal distribution
boptimization done by tuning three flow rates and honoring constraints
ccost of 3 ACS = $2.25 million
Each optimization yields a set of flow rates and is the optimum strategy to be employed at the corresponding confidence level. As the risk aversion factor (or confidence level) increases, both the gain and the risk decrease. For this reservoir, even at the 97.7% confidence level, the cost of the ACS is paid back in 2 years.
These results can be used in various ways. First, given the uncertainty in the width of the high-permeability zone, a decision as to whether to implement ACS may be made. An examination of the gains at various confidence levels shows whether the gain covers the investment. If it pays back at the 97.7% level, the technology is probably worth implementing. Table 2 thus provides a decision-making tool to evaluate the risk-reward of using this technology.
Second, the optimum flow rates point to an optimum operating strategy. Depending on the confidence level, the flow rates that are most likely to lead to that specific gain may be chosen, assuming no more analysis or measurements are performed to reduce the uncertainty in the width of the high-permeability zone. This leads to a third benefit: value of information (i.e., sensor value). The uncertainty in the gain achieved is due to the uncertainty in the reservoir property; i.e., the width of its high-permeability zone. Thus, any measurement that could reduce the uncertainty in this property will consequently reduce the uncertainty in the gain achieved or, at the same confidence level, increase the gain to be realized. This additional improvement in gain is obviously the value of the measurement used to obtain additional reservoir information (which will be described in more detail below with respect to value of information).
Stochastic reservoir optimization—two or more uncertain parameters: Now a second uncertainty, aquifer strength is introduced using the same example as above. As before, a normal distribution is assumed for both the variation in the width of this zone (mean of 700 ft and a standard deviation of 327 ft) and the aquifer PI (mean 4500 bpd/psi and standard deviation of 2858 bpd/psi). For the case of two uncertain parameters considered here, there will be N=n2 (m=2) equally probable scenarios, see equations (1) and (2) above. The gain in NPV due to ACS implementation is calculated for each of these equally probable scenarios. The mean μ (and standard deviation σ) of the NPV gain for the stochastic case is then the weighted NPV gain (and weighted standard deviation) of the N possible scenarios, as described by equations (1) and (2) above, where wi here is 1/N and gi is the gain in NPV for scenario i.
Because the reference case NPV is a function of two uncertain parameters, it is important, as explained above, to define the objective function as the gain in NPV rather than an absolute value.
Following the method proposed above, a stochastic optimization may be performed. The optimization is done using the weighted mean, μ and the weighted standard deviation, σ, of the NPV gains calculated for these N scenarios. An objective function may be constructed for this case by choosing to optimize a linear combination of the mean and standard deviation, as described by equation (3) above, where λ is a risk aversion factor chosen from a set of preferred risk aversion levels. As above, the objective function maximizes μ for a given σ or minimizes σ for a given μ. A value of λ equal to zero causes the optimization algorithm to maximize the mean without concern for the standard deviation. As the value of λ increases, the optimization algorithm puts more weight into minimizing the standard deviation, i.e., it reduces risk. If the gains in NPV of the various individual scenarios are assumed to be normally distributed, then the weighted mean corresponds to a 50% confidence level or a λ equal to zero in equation (3). When λ is equal to one, the function being optimized is (μ−σ), and the gain, assumed normally distributed, is maximized at an 84% confidence level. For the analysis here, three equally probable points for each uncertain variable (n=3) are used. Table 3 below summarizes the gains achieved at various confidence levels. Each optimization yields a set of flow rates and reflects the optimum strategy to be employed at the corresponding confidence level. As the risk aversion factor increases, the gain decreases but the spread in the gains also decreases. Thus, confidence about the result increases. For the reservoir of the present example, at the 84% confidence level, the cost of the ACS is paid back in two years but not at the 97.7% confidence level.
TABLE 3
Decision table for cost-benefit analysis of implementing
ACS technology at various confidence levels in the presence
of uncertainty in width of high permeability zone and aquifer PI.
Confidence
λ
Level*
Function optimized
Gain
Benefit>Cost**?
0
50%
F = μ − 0σ
$9.33 MM
Yes
1
84%
F = μ − 1σ
$4.88 MM
Yes
2
97.7%
F = μ − 2σ
$1.88 MM
No
*assumes normal distribution
**cost of three ACS is estimated to be $2.25 MM
These results can be used in various ways as described above. First, given the uncertainty in the two properties, a decision on the implementation of three ACS may be made. By examining the gains at the various preferred confidence levels, it can be determined whether the gain covers the investment. Table 3 thus provides a tool to evaluate the risk-reward of using this technology. Second, the optimum flow rates obtained for each λ point to an optimum operating strategy. Depending on the confidence level, the optimum flow rates that would ensure at least that gain may be identified (assuming that no more analysis or measurements are performed to reduce the uncertainty in the properties). This leads to the third application of this tool, which is the valuation of information (such as that collected by using sensors and making measurements) discussed below.
The above example considers only whether or not to implement a single strategy. The above example and decision tables may be repeated to account for one or more competing strategies and the gains compared at common risk aversion levels.
2. Valuation of Information (i.e., Sensor Value)
The risk factor in reservoir management arises due to uncertainty in physical and financial parameters associated with the reservoir. Minimizing the risk level involves gathering information about these parameters to reduce or eliminate the inherent uncertainties. Gathering information involves making a new measurement, implementing a new sensor, etc. Based on the physics of the measurement and the interpretation of the information, the uncertainty may be partially or completely eliminated. The cost of gathering this information should be balanced against the value the information brings to the stochastic optimization process through increased value of the goal being optimized (such as NPV). The quantification of the value of information is a valuable aid in designing measurement strategies to reduce uncertainties and, therefore, risk levels in reservoir management. When the uncertainty is completely eliminated, the value of information becomes the value of perfect information. This is demonstrated using the example of the reservoir with two uncertainties discussed above.
The decision table for the example (see Table 3) provides the associated risk level for each gain. The uncertainty in the realizable gain is due to the uncertainty in the reservoir properties—the width of its high permeability zone and the aquifer strength in the example introduced above. Thus, if the width is very small, then the gain due to the ACS is lower. To minimize the downside risk of the decision to implement ACS, one can optimize with a higher risk aversion factor and perform a cost-benefit analysis with a higher confidence. The next logical decision is to further minimize the risk associated with this decision by implementing a monitoring strategy that yields information about the reservoir properties, such as a logging suite or permanent sensors. Thus, any measurement that could reduce the uncertainty in these properties will consequently reduce the uncertainty in the gain achieved or, at the same confidence level, increase the gain to be realized. This additional improvement in gain is the value of the measurement or sensor used and is the value of information. The value of this information is the additional benefit it brings to reservoir management goals (either in terms of better NPV, higher hydrocarbon production, etc.) when the uncertainty in the property is reduced. By comparing the added benefit of a measurement with the cost of the measurement, a decision on performing the measurement may be made.
As shown in
Value of perfect information: If it is assumed that the measurements made completely eliminate all uncertainty, then the expected increase in gain attributable to those measurements is the value of perfect information. Accordingly, another embodiment of the invention as shown in
Using the example with two uncertainties described above, the methodology for estimating the value of perfect information is further described below.
Using the stochastic optimization, at the 84% confidence level (risk aversion factor of 1), the NPV of the same scenario is $10.4 MM. The additional increase in gain ($4.8 MM) over that chosen confidence level, i.e., the difference between these two values, is the potential benefit realized by completely eliminating the uncertainty in these properties. This is the maximum potential value of perfect information, such as from a sensor measurement. Because a stochastic model with nine equally probable scenarios is used, and there is no a priori knowledge of the true scenario, there is a distribution associated with this value (differences between the maximum gain and the 84% confidence level in
The value of perfect information is an upper limit to the benefit of a measurement in this case. In reality, the actual benefit would depend on the type of measurement made (i.e., sensor used), the ease of its interpretation, the time stage in production at which this information is obtained, and the extent to which the uncertainty is eliminated. Using the methodology outlined in
Further analysis may be performed to determine which of the many uncertain properties (two or more) is more important to the sensing process.
The reservoir risk management methodologies presented herein can help determine the benefit of implementing a specific technology or exploitation strategy and the gain to be realized by selecting a suitable monitoring strategy. Stochastic optimization using a linear combination of weighted mean gains and standard deviation of gains enables determination of gains from these technologies at various risk aversions levels. The technique has been demonstrated for valuation of ACS technologies in a reservoir with uncertainty in one or more properties. A decision-making table showing gains at various confidence levels facilitates cost-benefit analysis. The value of information is estimated by calculating the additional increase in gain that is due to partial or total elimination of uncertainty in the reservoir parameters. Such valuation is important in determining the merits of applying the technology and/or collecting information in a stochastic reservoir as well as in pricing them.
The invention has been described herein with reference to certain examples and embodiments. It will, however, be evident that various modifications and changes may be made to the embodiments described above without departing from the scope of the invention as claimed.
Raghuraman, Bhavani, Couet, Benoit
Patent | Priority | Assignee | Title |
10262280, | Jul 13 2015 | ConocoPhillips Company | Ensemble based decision making |
10318663, | Jan 26 2011 | ExxonMobil Upstream Research Company | Method of reservoir compartment analysis using topological structure in 3D earth model |
10584570, | Jun 10 2013 | ExxonMobil Upstream Research Company | Interactively planning a well site |
10961826, | Dec 19 2016 | ConocoPhillips Company | Subsurface modeler workflow and tool |
10989835, | Jan 25 2013 | Schlumberger Technology Corporation | Methods and systems for calculating and evaluating value of information for reservoir fluid models derived from DFA tool data |
11004026, | Nov 13 2017 | ADVANCED NEW TECHNOLOGIES CO , LTD | Method and apparatus for determining risk management decision-making critical values |
11340381, | Jul 02 2019 | Saudi Arabian Oil Company | Systems and methods to validate petrophysical models using reservoir simulations |
11598185, | Nov 24 2014 | Schlumberger Technology Corporation | Methods for adaptive optimization of enhanced oil recovery performance under uncertainty |
7725302, | Dec 02 2003 | Schlumberger Technology Corporation | Method and system and program storage device for generating an SWPM-MDT workflow in response to a user objective and executing the workflow to produce a reservoir response model |
7813935, | Jan 13 2004 | WEATHERFORD TECHNOLOGY HOLDINGS, LLC | System for evaluating over and underbalanced drilling operations |
7966166, | Apr 18 2008 | Schlumberger Technology Corporation | Method for determining a set of net present values to influence the drilling of a wellbore and increase production |
8099266, | Sep 03 2004 | Drilling Systems Ltd | Method and system for the design of an oil well |
8122965, | Dec 08 2006 | HORTON WISON DEEPWATER, INC | Methods for development of an offshore oil and gas field |
8332195, | Nov 17 2008 | Landmark Graphics Corporation | Systems and methods for running a multi-thread simulation |
8352226, | Jan 31 2006 | Landmark Graphics Corporation | Methods, systems, and computer-readable media for real-time oil and gas field production optimization using a proxy simulator |
8352227, | Oct 30 2006 | Schlumberger Technology Corporation | System and method for performing oilfield simulation operations |
8392164, | Aug 06 2007 | IFP | Method for evaluating an underground reservoir production scheme taking account of uncertainties |
8457997, | Apr 29 2005 | Landmark Graphics Corporation | Optimization of decisions regarding multiple assets in the presence of various underlying uncertainties |
8469090, | Dec 01 2009 | Schlumberger Technology Corporation | Method for monitoring hydrocarbon production |
8504341, | Jan 31 2006 | Landmark Graphics Corporation | Methods, systems, and computer readable media for fast updating of oil and gas field production models with physical and proxy simulators |
8527248, | Apr 18 2008 | WESTERNGECO L L C | System and method for performing an adaptive drilling operation |
8548785, | Apr 27 2009 | Schlumberger Technology Corporation | Method for uncertainty quantification in the performance and risk assessment of a carbon dioxide storage site |
8793111, | Jan 20 2009 | Schlumberger Technology Corporation | Automated field development planning |
8818777, | Oct 30 2006 | Schlumberger Technology Corporation | System and method for performing oilfield simulation operations |
8849623, | Dec 16 2008 | ExxonMobil Upstream Research Company | Systems and methods for reservoir development and management optimization |
8884964, | Apr 22 2008 | ExxonMobil Upstream Research Company | Functional-based knowledge analysis in a 2D and 3D visual environment |
8892407, | Oct 01 2008 | ExxonMobil Upstream Research Company | Robust well trajectory planning |
8931580, | Feb 03 2010 | ExxonMobil Upstream Research Company | Method for using dynamic target region for well path/drill center optimization |
9026417, | Dec 13 2007 | ExxonMobil Upstream Research Company | Iterative reservoir surveillance |
9128693, | Nov 10 2007 | Landmark Graphics Corporation | Systems and methods for workflow automation, adaptation and integration |
9223594, | Jul 01 2011 | ExxonMobil Upstream Research Company | Plug-in installer framework |
9228415, | Oct 06 2008 | Schlumberger Technology Corporation | Multidimensional data repository for modeling oilfield operations |
9593558, | Aug 24 2010 | ExxonMobil Upstream Research Company | System and method for planning a well path |
9595129, | May 08 2012 | ExxonMobil Upstream Research Company | Canvas control for 3D data volume processing |
9658674, | Mar 11 2013 | Samsung Electronics Co., Ltd. | Mobile system optimization method |
9727928, | Mar 14 2013 | National Technology & Engineering Solutions of Sandia, LLC | Financial methods for waterflooding injectate design |
9864098, | Sep 30 2013 | ExxonMobil Upstream Research Company | Method and system of interactive drill center and well planning evaluation and optimization |
9874648, | Feb 21 2011 | ExxonMobil Upstream Research Company | Reservoir connectivity analysis in a 3D earth model |
9940414, | Feb 24 2014 | Landmark Graphics Corporation | Total asset modeling with integrated asset models and persistent asset models |
Patent | Priority | Assignee | Title |
5148365, | Aug 15 1989 | International Business Machines Corporation | Scenario optimization |
5924048, | Mar 14 1997 | Atlantic Richfield Corporation | Automated material balance system for hydrocarbon reservoirs using a genetic procedure |
5992519, | Sep 29 1997 | Schlumberger Technology Corporation | Real time monitoring and control of downhole reservoirs |
6128579, | Mar 14 1997 | Atlantic Richfield Corporation | Automated material balance system for hydrocarbon reservoirs using a genetic procedure |
6182014, | Nov 20 1998 | Schlumberger Technology Corporation | Method and system for optimizing logistical operations in land seismic surveys |
6205402, | Mar 20 1998 | Institut Francais du Petrole | Method for automatically forming a model simulating the stratigraphic structure of an underground zone |
6895383, | Mar 29 2001 | Accenture Global Services Limited | Overall risk in a system |
6980940, | Feb 22 2000 | Schlumberger Technology Corp. | Intergrated reservoir optimization |
7006992, | Apr 06 2000 | PACKWOOD, THOMAS R | Risk assessment and management system |
20040215551, | |||
20050071266, |
Executed on | Assignor | Assignee | Conveyance | Frame | Reel | Doc |
Dec 23 2002 | Schlumberger Technology Corporation | (assignment on the face of the patent) | / | |||
Jan 02 2003 | RAGHURAMAN, BHAVANI | Schlumberger Technology Corporation | ASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS | 013901 | /0323 | |
Jan 02 2003 | COUET, BENOIT | Schlumberger Technology Corporation | ASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS | 013901 | /0323 |
Date | Maintenance Fee Events |
Aug 29 2012 | M1551: Payment of Maintenance Fee, 4th Year, Large Entity. |
Sep 15 2016 | M1552: Payment of Maintenance Fee, 8th Year, Large Entity. |
Sep 17 2020 | M1553: Payment of Maintenance Fee, 12th Year, Large Entity. |
Date | Maintenance Schedule |
Mar 31 2012 | 4 years fee payment window open |
Oct 01 2012 | 6 months grace period start (w surcharge) |
Mar 31 2013 | patent expiry (for year 4) |
Mar 31 2015 | 2 years to revive unintentionally abandoned end. (for year 4) |
Mar 31 2016 | 8 years fee payment window open |
Oct 01 2016 | 6 months grace period start (w surcharge) |
Mar 31 2017 | patent expiry (for year 8) |
Mar 31 2019 | 2 years to revive unintentionally abandoned end. (for year 8) |
Mar 31 2020 | 12 years fee payment window open |
Oct 01 2020 | 6 months grace period start (w surcharge) |
Mar 31 2021 | patent expiry (for year 12) |
Mar 31 2023 | 2 years to revive unintentionally abandoned end. (for year 12) |